CSCE 978 (Spring 2006) Homework 1

Assigned Friday, January 27
Due Sunday, February 12 at 11:59 p.m.
Total points: 115

Submit a single .pdf file with your writeup of the results for all the homework problems, including the last problem. Only pdf will be accepted, and you should only submit one pdf file, with the name username.pdf, where username is your username on cse.

Submit everything by the due date and time using the web-based handin program.

On this homework, you must work on your own and submit your own results written in your own words.

1. (20 pts) Choose any algorithm (except for support vector machines) that could benefit from the kernel trick. In what way is a dot product used in this algorithm, and how might a kernel help? Are there any drawbacks to using a kernel in this case? What kernel do you think would work well and why? Remember to cite relevant papers and other sources in your report.

2. (40 pts) You may not have heard, but February is Adopt-A-Kernel-Month. In honor of this, you are to choose a kernel that you will use for the rest of the semester. Read about it, get to know it, understand it, love it, and write about it. In particular, you should discuss where it originated, on what problems it has been applied, and contrast it with other kernels on similar problems (i.e. how successful was your kernel versus others). What is its corresponding remapping? If that is not known, then discuss the intuition behind using it as a dot product/similarity measure. Does your kernel have other properties that make it interesting and useful? How applicable would it be for use in kernelizing the algorithm you discuss in Problem 1? Make sure you cite relevant papers and other sources in your report, and note that you may not use a kernel that we've covered in class so far (and not a simple modification, combination, etc. of one we've covered in class so far).

3. (20 pts) Do Exercise 2.18 on page 57 of the text.

4. (10 pts) Do Exercise 5.6 on page 147 of the text.

5. (15 pts) Do Exercise 5.8 on page 148 of the text.

6. (5 pts) List at least three topics that you feel would be good subjects for a class project. These topics may be on any material related to SVMs or other kernel methods, i.e. they need not be related to topics we have covered in class so far.

7. (5 pts) State the approximate amount of time you spent on each problem.

Last modified 16 August 2011.